U.S. patent number 9,922,428 [Application Number 15/241,547] was granted by the patent office on 2018-03-20 for vibration image acquisition and processing.
This patent grant is currently assigned to CRYSTAL INSTRUMENTS CORPORATION. The grantee listed for this patent is Crystal Instruments Corporation. Invention is credited to Hongjian Gao, Yihao Liu, Weijie Zhao, James Zhuge.
United States Patent |
9,922,428 |
Liu , et al. |
March 20, 2018 |
Vibration image acquisition and processing
Abstract
A shaker test apparatus is provided along with a method of
collecting and processing images, wherein a shaker table is driving
the device under test by a vibration controller at a known
vibration frequency and period, wherein a device under test obtains
a steady-state vibration characteristic of that excitation
frequency when mounted on the shaker table. While the device under
test is being excited, a trigger signal controller triggers a
camera to capture a series of still image frames at a regular
sampling frequency that is less than the vibration frequency
(under-sampling), and a timer associated with the camera records a
timestamp of an image capture time for each image frame. A computer
processor uses the timestamps to remap the order of the image
frames, shifting each frame's capture time backwards by a specified
multiple of vibration periods in order to correctly represent a
single vibration period beginning with an earliest captured
image.
Inventors: |
Liu; Yihao (Hangzhou,
CN), Gao; Hongjian (Linhai, CN), Zhuge;
James (Palo Alto, CA), Zhao; Weijie (Los Gatos, CA) |
Applicant: |
Name |
City |
State |
Country |
Type |
Crystal Instruments Corporation |
Santa Clara |
CA |
US |
|
|
Assignee: |
CRYSTAL INSTRUMENTS CORPORATION
(Santa Clara, CA)
|
Family
ID: |
61191995 |
Appl.
No.: |
15/241,547 |
Filed: |
August 19, 2016 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20180053310 A1 |
Feb 22, 2018 |
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G01M
7/022 (20130101); H04N 13/296 (20180501); G06T
7/292 (20170101); H04N 13/239 (20180501); G01M
7/025 (20130101) |
Current International
Class: |
G06K
7/00 (20060101); G06T 7/20 (20170101); B06B
3/00 (20060101); G01H 17/00 (20060101); H04N
5/228 (20060101); G01N 29/34 (20060101); G01H
1/00 (20060101) |
Field of
Search: |
;382/151,154,173,312
;348/208.99,373,374,208.4,208.1,208.2,208.6,169,208.11,208.7,335
;73/570,662,663,668,664,667,579,582,577 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Chawan; Sheela C
Attorney, Agent or Firm: Schneck; Thomas
Claims
What is claimed is:
1. A method of collecting and processing images of a device under
test on a shaker table and generating a model of 3-dimensional
movement of the device under test, comprising: driving a shaker
table with sinusoidal signal at a known exciting frequency, a
device under test mounted to the shaker table obtaining a
steady-state vibration characteristic corresponding to the
excitation frequency; imaging the device under test, while the
shaker table is driving the device under test, as a series of still
image frames taken by more than one camera at a regular sampling
frequency that is less than the shaker excitation frequency, each
image frame being stored together with a timestamp recording an
image capture time for that frame; reordering the stored series of
still image frames from each camera, using the timestamps, into
corresponding remapped series of images of each camera that
represent a single vibration period beginning with an earliest
captured image, where the reordering shifts each image frame, other
than the earliest image frame, backwards in time by a specified
multiple of vibration periods; and creating a model, using the
reordered series of images from more than one camera, of
3-dimensional movement of the device under test.
2. The method as in claim 1, wherein an image capture rate of least
one camera is not more than 200 Hz.
3. The method as in claim 1, wherein at least one camera is a
stereo camera.
4. The method as in claim 1, wherein the timestamps are generated
by an interface card timer associated with the camera at the close
of the camera's shutter.
5. The method as in claim 1, wherein the timestamps are generated
by a system clock of a computer receiving and storing the image
frames.
6. The method as in claim 1, wherein the shaker table is driven at
a series of stepped vibration frequencies, each frequency step
lasting for a specified number of vibration cycles sufficient to
capture a set of at least two dozen image frames.
7. The method as in claim 1, further comprising calculating a
resonant frequency of the device under test from the 3-dimensional
movement modeled by the images.
8. The method as in claim 1, further comprising calculating a mode
shape of the device under test from the 3-dimensional movement
modeled by the images.
9. The method as in claim 1, wherein the shutters of all cameras
are controlled and synchronized by one timer source.
10. The method as in claim 1, wherein multiple shakers are employed
to drive the device under test at the same time at same excitation
frequency.
11. A shaker test apparatus, comprising: a shaker table; a
vibration controller connected to the shaker table so as to drive
the shaker table at a known frequency and period, wherein a device
under test obtains a steady-state shake characteristic of that
excitation frequency when mounted on the shaker table; more than
one camera directed toward the shaker table with any device under
test mounted on that shaker table; a trigger signal controller
coupled to the camera so as to trigger the camera, while the shaker
table is driving the device under test, to capture a series of
still image frames at a regulated sampling frequency that is less
than the shaking frequency; a timer coupled to the camera(s) so as
to record a timestamp of an image capture time for each image
frame; and a computer processing algorithm for reordering the
stored series of still image frames, using the timestamps, into a
remapped series of the images representing a single vibration
period beginning with an earliest captured image, where the
reordering shifts each image frame, other than the earliest image
frame, backwards in time by a specified multiple of shake periods.
Description
TECHNICAL FIELD
The present invention relates to vision-based vibration testing of
devices under test by means of a shaker table or similar vibration
test equipment, and in particular relates to the acquisition and
processing of images of the device under test while undergoing such
vibration testing.
BACKGROUND ART
In vibration testing, including modal analysis, a large amount of
hardware equipment would be required for the large scale model
testing when traditional vibration sensors are used. Further time
and efforts of a group of engineers would be significant too in
order to carry out the tests over hundreds of measurement points on
the device under test.
Since a few decades ago, DIC technology emerged with the
development of the high speed camera. In recent decades, the
advances of 3D stereo camera and associated DIC methods make it
possible to measure the 3-dimensional vibration deformation of the
device under test. This technology results in the full field
measurement of the device under test, and which also can be done
with significant less time and effort.
In a conventional DIC based vibration testing system, a shaker
table is used to drive the device under test, while one or more
cameras image the device under test being excited. Because there is
typically no synchronization between the vibration of the shaker
table and the exposure trigger for the cameras, expensive
high-speed cameras need to be used to ensure an image sampling rate
that is at least twice the vibration frequency of the device under
test. This is the normal sampling mode for image acquisition in
such systems.
As a possible cheaper alternative, one might contemplate use of
low-cost, but also low-speed, cameras with some kind of
under-sampling technique by means of synchronized triggering of
such cameras. However, the existence of jitter (time deviations) in
either the trigger signal itself, or more commonly in the cameras'
response, severely limits the potential accuracy of such a scheme.
Any random deviations in the cameras' frame capture times will
result in errors during subsequent processing to reconstruct a
correct sequence of image frames.
SUMMARY DISCLOSURE
A shaker test apparatus and method acquires and processes a series
of images of a device under test on a shaker table in order to
capture and analyze the vibration of the device under test when
excited at a known frequency. In particular, timestamping of the
captured images is used to facilitate under-sampling and then
remapping of the sequence of images so that a low trigger rate for
one or more cameras can be employed even in the presence of a much
higher excitation frequency of the vibration.
The shaker test apparatus comprises a shaker table with a vibration
controller, one or more cameras (any of which could be stereo
cameras) with a trigger signal controller and timer, and a
processing computer. The vibration controller is connected to the
shaker table so as to drive the shake table at a known vibration
frequency, and period. When a device under test is mounted on the
shaker table and driven, the device under test will settle to a
steady-state vibration characteristic that can be captured by a
series of images. The camera(s) is directed toward the device under
test so as to image the device under test that mounted on that
shaker table. The trigger signal controller is coupled to the
camera so as to trigger, while the shaker table is driving the
device under test, the capture by the camera of a series of still
image frames at a regular sampling frequency that is less than the
excitation frequency, thereby under-sampling the vibration.
However, the timer that is coupled to the camera records a
timestamp of the image capture time for each image frame. When
multiple cameras are used, all cameras should be synchronized with
a common trigger signal controller for comparable timestamps of
related image frames of the different camera views.
Using the timestamps associated with each frame, a processing
algorithm running on the computer reorders the stored series of
still image frames from each camera into corresponding remapped
series of those same images so as to represent a single vibration
period. In particular, the earliest captured image frame is used to
represent the start of a vibration period and its timestamp serves
as the reference time. The known vibration period from the
vibration controller is added to the reference time to obtain the
end time of one single vibration period. For each captured image
frame other than that earliest image frame, the reordering process
takes that frame's associated timestamp and shifts its capture
backwards in time by a multiple of vibration periods, until an
adjusted capture time falls within the vibration period. Once all
of the time shifts are completed, the frames are put into a new
"remapped" order using the adjusted capture times.
The reordered image sets from the various cameras can be used by
available analysis tools to model 3-dimensional movement of the
device under test, which can be used to calculate resonant
frequency, mode shapes, displacement amounts, and other
parameters.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a schematic block diagram of a shaker test apparatus in
accord with the present invention.
FIGS. 2A to 2C are graphs of vibration displacement versus time
that illustrate the reconstruction of an original signal from
under-sampled data. In FIG. 2B, the under-sampling is not taken
into account, resulting in incorrect reconstruction, whereas in
FIG. 2C, the under-sampled data is correctly remapped.
FIGS. 3A and 3B are graphs of reconstructed vibration displacement,
first without and then with the use of timestamps, thereby
illustrating the benefit of timestamps for accurate
reconstruction.
DETAILED DESCRIPTION
With reference to FIG. 1, a shaker test apparatus includes a set of
vibration test equipment 10 (including a shaker 11, a vibration
controller 13, and power amplifier 15) for a device under test 17,
and a set of image acquisition equipment 20 (including one or more
cameras 21 with an associated timer 25, a trigger signal controller
23 and a computer 27). The vibration controller 13 outputs an
excitation signal 14 to the amplifier 15 based on a configuration
of user-specified parameters, including signal type of Sine,
amplitude, frequency, etc. The power amplifier 15 will amplify the
excitation signal 14 to drive the shaker 11. The fundamental
frequency of the excitation signal 14 from the controller 13 to the
shaker 11 can be either fixed, or stepped through multiple fixed
frequencies, or slowly swept over a certain frequency range. It is
important for subsequent processing of acquired images both that
the excitation frequency be known and also that a device under test
17 mounted on the shaker 11 be allowed to obtain a steady-state
motion stage. This means the transient responses of the device
under test from the initial applying of the excitation frequency
will decay and settle. A steady-state motion is normally achieved
after probably several vibration cycles at the excitation
frequency, so there is a bit of delay before beginning image
capture by the camera.
The camera 21, which can preferably be a stereo camera, receives a
trigger signal 24 from the trigger signal controller 23. The form
of trigger signal 24 from the controller 23 can vary, but the most
often used trigger signal is an electrical pulse signal with
accurate rising or falling edge. The camera 21 will operate its
shutter based on the received trigger signal 24 and as a
consequence captures images 26 at either the rising or falling edge
of the trigger signal 24. The characteristics of trigger timing
from the controller 23 can be set according to user specifications
within the capabilities of the camera. In order to permit low speed
cameras to be used, an under-sampling technique is used wherein the
trigger rate can be at a mere fraction of the shaker's excitation
frequency. For definitional purposes, high-speed vibration is any
frequency in excess of 200 Hz, and likewise, high-speed imaging is
any image capture rate in excess of 500 Hz. Low-speed cameras
operate below that limit, typically 30 or 60 frames/sec, and in
case of the present invention can be applied to the imaging of
devices under test being driven at any frequency, high or low. In
any case, very-high-speed vibration over about 1 KHz tends to
produce very limited displacements (under 0.2 mm) in the device
under test which are difficult to image with any camera.
Under-sampling refers to any image capture rate of the camera 21
that is less than twice the vibration frequency of the shaker 11.
Nyquist-Shannon sampling theorem normally dictates a sample rate
for any data acquisition that is at least twice the observed
frequency. However, if a signal being analyzed has a limited
bandwidth which is exactly known, then it becomes possible to
sample at a lower frequency, i.e. to under-sample the signal, and
remap those sample results so as to reconstruct the original
signal. To illustrate this, we refer to an example given in FIGS.
2A-2C wherein an original signal 31 is a sine-wave with a known
frequency of 100 Hz. If the signal is sampled at a frequency below
200 Hz, say 80 Hz, there will be a set of sample results, marked by
the black dots 33a-33e (at every 121/2 milliseconds in this
example). Normally, if one were not to take into account the
under-sampling, the signal reconstruction will result in aliasing,
as shown in FIG. 2B, where the reconstructed signal 35 with
corresponding sample points 37a-37e every 121/2 milliseconds is
incorrect. Instead of recreating the original 100 Hz sine-wave, a
50 Hz sine-wave is constructed. However, with knowledge of the
signal frequency and the under-sampling rate, it is possible to
remap the time of sample results into time-adjusted sample points
43a-43e and thereby correctly reconstruct the signal 41, as seen in
FIG. 2C.
This illustration of the theory underlying an under-sampling
technique assumes so far that there is no jitter (random
fluctuation) in the sample acquisition times. The existence of any
such jitter will, unless accounted for, result in errors in
remapping the sample points and thus reduced accuracy in the
reconstructed signal. Successful application of under-sampling
requires high accuracy of the trigger signal controller 23. In
particular, the time resolution of the trigger signal controller 23
should be at least higher than the frequency range of the
excitation signal 14 driving the shaker 11. For example, if the
excitation signal 14 to the shaker 11 is in the range of 1 KHz (1
ms period), the time accuracy of the camera shutter's trigger
signal 24 should be in a few microsecond range or better. Even so,
time deviations between the signal 24 from trigger signal
controller 23 and the capture of the image 26 by camera 21 will
decrease the accuracy of signal reconstruction.
Accordingly, the present invention introduces the use of
time-stamping to eliminate the influence of jitter. The exact
capture time for each image frame 26 is recorded so that even if
there is jitter, it can be correctly accounted for in the
reconstruction process. Timestamps can be generated from a system
clock in the computer 27 receiving the images 26, or more
preferably by an interface card timer 25 associated with the camera
21. It is recommended for time-stamping of each image frame to take
the interface card timer at the time that the camera shutter is
closed, for best accuracy, and to embed that capture time in the
image file.
Since jitter within a interface card timer 25 is less than 500 ns
(typically 120 ns peak-to-peak jitter and 40 ns repeat jitter), the
generated timestamp when using such a timer has a higher timing
resolution than most high-timing trigger signal controllers 23, and
therefore provided better results when performing under-sampling.
As seen in FIG. 3A, for a shaker table operating at 5 Hz and a
stereo camera triggered at 1.242 Hz, the under-sampling without the
timestamp technology will have an adj. R-square value of 0.92498
when fitting the acquired data to a 5 Hz sinewave. But it can be
seen in FIG. 3B, that under the same conditions while using the
timestamps, the fitted data will have a much better adj. R-square
value of 0.97737.
Capture of timestamped images 26 by the camera or cameras 21
generates a set of image frames which are stored by the computer
27. Using the timestamps associated with each frame, the processing
algorithm running on computer 27 reorders the stored series of
still image frames for each camera into a corresponding remapped
series of those same images so as to represent a single vibration
period. In particular, the earliest captured image frame is used to
represent the start of a vibration period and its timestamp serves
as the reference time. The known vibration period from the
vibration controller is added to the reference time to obtain the
end time of one single shake period. For each captured image frame
other than that earliest image frame, the reordering process takes
that frame's associated timestamp and shifts its capture backwards
in time by a multiple of one or more shake periods (basically
modular arithmetic by means e.g. of successive subtractions from
the timestamped image capture time), until an adjusted capture time
falls within the shake period. Thus, for example, if the known
shake frequency is 100 Hz, for a shake period of 10 ms, then 10 ms
will be subtracted one or more times from the frame's capture time
to obtain an adjusted timestamp value that is within 10 ms of the
first frame's timestamp. Once all of the time shifts are completed,
the frames are put into a new "remapped" order using the adjusted
capture times.
Other image sets from additional cameras with different points of
view are similarly treated. Preferably, the shutters of all cameras
are controlled and synchronized by one and the same timer source so
that timestamps for the multiple image sets will be comparable for
easier modeling of the 3-dimensional movement of the device under
test by known analytical software tools.
Having been put into a remapped order, the image sets can then be
played as a moving picture or analyzed to determine degree of
displacement and other parameters of the device under test. For
example, the reordered image sets from each camera can be used to
create model of 3-dimensional movement of the device under test by
means of presently available analytical software tools, from which
information such as resonant frequency or mode shape can be then be
calculated.
* * * * *